Most memory systems utilize a multi-drop connection between the memory controller and multiple memory devices to increase the memory density on the system board and at the package level. However, this approach does not guarantee good signal integrity and enough timing margin in high speed applications, for example, over 100 MHz frequency. Therefore, there is a need for an alternative memory system architecture that achieves acceptable signal integrity for high speed applications. Such an alternative has been found in architectures that feature multiple devices that are serially interconnected. In such architecture, a memory controller is connected to a first memory device with a link, and that memory device is connected to a next memory device with another link and so on. The use of links between devices overcomes some of the problems with signal integrity for high speed applications. However, unlike a multi-drop architecture, where specific devices can be activated while remaining devices are passive, all of the devices in a serially interconnected architecture are active because they must be available to pass signals on to the next device in the serial interconnection.